Damage Detection in Four Point Bending Test on Benchmark RC Structure Using Feature based Fusion

Abstract

Sensor fusion has attracted significant attention in recent years due to its usability for many applications in real life. This encourages to use the fusion technique for damage detection in concrete structures. The major diffculty in a concrete damage detection lies in the fact of early crack detection and take proper action before the propagation of cracks. Therefore, different techniques were performed on a benchmark RC beam, which was subjected to four point loading. Then the features from multiple sensors were fused for early crack detection. In this framework, we first represent each measurement technique separately, in which coefficient of peak to peak amplitude from ultrasonic measurement, and the strain measured by strain gauges serve as the features indicator for damage detection. Canonical correlation analysis (CCA) is then applied to both features to construct a combination of features of peak to peak and strain matrices. The result indicates the possibility of using a feature-based fusion algorithm more robustly, and it increases the damage detection probability by reducing false alarm ratio.